U.S. patent number 6,876,455 [Application Number 10/210,921] was granted by the patent office on 2005-04-05 for method and apparatus for broadband optical end point determination for in-situ film thickness measurement.
This patent grant is currently assigned to Lam Research Corporation. Invention is credited to Vladimir Katz, Bella Mitchell.
United States Patent |
6,876,455 |
Katz , et al. |
April 5, 2005 |
Method and apparatus for broadband optical end point determination
for in-situ film thickness measurement
Abstract
A method for determining a film thickness in a semiconductor
substrate is provided. The method initiates with providing multiple
layers on the semiconductor substrate. Then, two reflectance
spectra are generated where each of the two reflectance spectra are
associated with different time periods while an upper layer is
being removed. Next, a difference between the two reflectance
spectra is calculated. Then, a curve is defined from the difference
between the two reflectance spectra. Next, the defined curve is
fitted by a known parametric function to determine the film
thickness. An endpoint detector and a CMP system are also
provided.
Inventors: |
Katz; Vladimir (Fremont,
CA), Mitchell; Bella (Antioch, CA) |
Assignee: |
Lam Research Corporation
(Fremont, CA)
|
Family
ID: |
34375046 |
Appl.
No.: |
10/210,921 |
Filed: |
August 1, 2002 |
Current U.S.
Class: |
356/503 |
Current CPC
Class: |
G01B
11/0625 (20130101) |
Current International
Class: |
G01B
11/06 (20060101); G01B 011/02 () |
Field of
Search: |
;356/485,492,503,504,497,498 ;250/599.27 |
References Cited
[Referenced By]
U.S. Patent Documents
Primary Examiner: Turner; Samuel A.
Attorney, Agent or Firm: Martine Penilla & Gencarella,
LLP
Claims
What is claimed is:
1. A method for determining a film thickness in a semiconductor
substrate, comprising: providing multiple layers on the
semiconductor substrate; generating two reflectance spectra, each
of the two reflectance spectra associated with different time
periods; calculating a difference between the two reflectance
spectra; defining a curve from the difference between the two
reflectance spectra; and fitting the defined curve to a best
fitting solution to determine the film thickness.
2. The method of claim 1, wherein the defined curve and the best
fitting solution are of a sinusoidal shape.
3. The method of claim 1, wherein the method operation of fitting
the defined curve to a best fitting solution to determine the film
thickness further includes: determining a best fit between the
defined curve and the best fitting solution based on a slope
fitting criterion.
4. The method of claim 1, further including: determining a trench
depth of a trench within the semiconductor substrate with one of
the two reflectance spectra.
5. The method of claim 4, wherein the method operation of
determining a trench depth of a trench within the semiconductor
substrate with one of the two reflectance spectra includes:
defining a curve from the one of the two reflectance spectra; and
fitting the curve to a found cosine based curve to determine the
trench depth.
6. The method of claim 5, wherein the cosine based curve is defined
by a cosine function, the cosine function being cos (2.pi.*2n.sub.1
*d.sub.0 *v), wherein n.sub.1 is an index of refraction for a first
layer, d.sub.0 is a trench depth, and v is a wave number.
7. The method of claim 1, wherein the multiple layers include a
silicon oxide layer disposed over a silicon nitride layer, the
silicon nitride layer disposed over a silicon substrate.
8. The method of claim 1, wherein the two reflectance spectra are
generated during one of a chemical mechanical planarization
operation and an etch operation.
9. The method of claim 1, wherein the two reflectance spectra
includes a dominant reflectance spectra reflected from a trench of
a silicon substrate.
10. The method of claim 1, wherein each reflectance spectra is
associated with an intensity.
11. The method of claim 1, wherein the film thickness is less than
about 600 nanometers.
12. An endpoint detection system, comprising: a detector including;
an emitter for generating light pulses having an intensity at
periodic time periods; a receiver for measuring a reflected light
spectrum from two of the light pulses; and a controller in
communication with the detector, the controller configured to
generate a curve for the two reflected light spectrum, the
controller configured to calculate a difference between the two
reflected light spectrum, the controller further configured to
define a curve from the difference between the two reflected
light.
13. The end point detection system of claim 12, wherein the
controller is further configured to execute a best fitting
algorithm to generate the curve corresponding to the difference
between the two reflected light spectrum.
14. The end point detection system of claim 12, wherein the
reflected light spectrum is reflected from a surface of a
semiconductor substrate having multiple layers disposed
thereon.
15. A chemical mechanical planarization (CMP) system, the system
comprising: a polishing surface having a window defined
therethrough; a carrier configured to support a semiconductor
substrate, the carrier further configured to force a surface of the
semiconductor substrate against the polishing surface; an endpoint
detector disposed on an underside of the polishing surface, the
endpoint detector configured to receive a reflected light signal
from the surface of the semiconductor substrate, the endpoint
detector including; an emitter for periodically generating an
intensity of light; a receiver for measuring a reflected light
spectrum from each intensity of light periodically generated; and a
controller in communication with the end point detector, the
controller configured to generate a curve by calculating a
difference between the reflected light spectrum from a first
generated intensity of light and a second generated intensity of
light and applying a best fitting algorithm to the difference.
16. The CMP system of claim 15, wherein the controller is further
configured to terminate a CMP process when a predetermined
thickness of the semiconductor substrate is obtained.
17. The CMP system of claim 15, wherein the controller determines
which of a plurality of found curves corresponds most closely to
the curve.
18. The CMP system of claim 15, wherein the function approximating
the sinusoidal based curve is represented as ##EQU5##
wherein n.sub.1 an index of refraction for a first layer, d.sub.1
represents a thickness of the first surface, .alpha..sub.j =n.sub.j
(d.sub.j -d.sub.1), and v is a wave number.
19. The end point detection system of claim 12, wherein the two of
the light pulses are successive light pulses.
Description
BACKGROUND OF THE INVENTION
The invention relates generally to semiconductor fabrication and
more specifically to in-situ film thickness measurement for thin
films deposited on a semiconductor substrate utilizing broad band
spectrometry.
During semiconductor fabrication there are multiple steps where an
underlying substrate is subjected to the formation and removal of
various layers. The small feature sizes and tight surface planarity
requirements, combined with the constant quest to increase
throughput, makes it highly desirable to stop the process when the
correct thickness has been achieved, i.e., when an endpoint has
been obtained for the process step.
Present optical end point detection (EPD) methods make use of Broad
Band Visible Spectra Spectrometry. The light of a lamp flash, i.e.,
shot, is returned from a wafer while an upper film is removed
through a chemical mechanical planarization (CMP) or etch process.
The light spectrum returned is measured for each shot and analyzed
according to a programmed algorithm. For many multilayer
transparent thin film structures, such as films associated with
shallow trench isolation (STI) applications, the reflectance from
the silicon substrate, is much greater then from upper layer film
interfaces. Accordingly, the underlying silicon provides the main
contribution in observed reflectance spectra. For instance, for a
STI structure the reflectance from the silicon (Si) substrate
surface, created at prior technological step trenches provides the
main contribution in total reflectance. As a result, the
reflectance spectra changes related to the layer of interest, i.e.,
the upper layer, are relatively small from shot to shot even at the
moment when upper layer is fully removed. Therefore, EPD methods
based on detecting the transition from one layer to the next
underlying layer are not efficient enough when dealing with these
thin transparent films encountered during upper layer removal
processes. In such cases the method of direct measurement of the
film thickness based on broadband visible spectrometry is
preferred. For transparent films the measured spectra are the
result of complex interference of light reflected from the wafer
surface and deeper layers of the wafer, and the light scattered
back from interlayer material. To obtain film thickness from these
complicated spectra, Fourier transform and simulation spectra
methods are typically used. One skilled in the art will appreciate
that Fourier transform expends the complex spectrum on periodic
terms. Peak locations in Fourier transform spectrum allow to
restore film thickness of the layers in a stack. This method is
applicable to relatively thick films. The periodic term
corresponding to the film thickness should have enough repetitions
in the measured spectrum so that peaks will be well separated and
distinguished. However, when the spectrometer measurement band is
300 nanometers-700 nanometers, this method is inadequate for
measuring the thickness of a layer that is less then about 300-600
of a nanometer (nm).
The simulation spectra methods are based on a multiparametric model
where the thickness of layers, reflectance of interlayer boundaries
and other characteristics of the structure taken as parameters. The
multiparametric model disadvantages include high sensitivity of the
approximation to precise values of introduced parameters and
existence of numeral, very close fittings with completely different
sets of fitting parameters. This uncertainty, increases with the
number of terms and parameters in the fitting model. In addition,
the above described methods are ineffective for layer thickness
measurement if reflectance of the layer surfaces is only small part
of total reflectance, such as the upper layer in an STI structure.
That is, the thickness of the layer can not be extracted from such
a spectrum because of the dominance of the reflectance from silicon
and multiple unknown parameters.
In view of the foregoing, there is a need to provide a method and
system to measure the thickness of layers deposited on a
semiconductor substrate through broadband visible spectrometry
irrespective of the contribution of the reflectance of the layer
surface to the total reflectance.
SUMMARY OF THE INVENTION
Broadly speaking, the present invention fills these needs by
providing a method and system that considers the differences
between multiple shots to discard reflectance spectra contribution
that does not change from shot to shot to reduce the number of
parameters. It should be appreciated that the present invention can
be implemented in numerous ways, including as an apparatus, a
system, a device, or a method. Several inventive embodiments of the
present invention are described below.
In accordance with one embodiment of the present invention, a
method for determining a film thickness in a semiconductor
substrate is provided. The method initiates with providing multiple
layers on the semiconductor substrate. Then, two reflectance
spectra are generated where each of the two reflectance spectra are
associated with different time periods. Next, a difference between
the two reflectance spectra is calculated. Then, a curve is defined
from the difference between the two reflectance spectra. Next, the
defined curve is fitted to a best fitting solution to determine the
film thickness.
In another embodiment, an endpoint detection system is provided.
The endpoint detection system includes a detector. The detector
includes an emitter for generating an intensity of light and a
receiver for measuring a reflected light spectrum from the
generated intensity of light. The endpoint detection system also
includes a controller in communication with the detector. The
controller is configured to generate a curve corresponding to the
reflected light spectrum. Additionally, the controller is
configured to execute a best fitting algorithm to generate a curve
corresponding to the reflected light spectrum.
In accordance with yet another embodiment of the present invention,
a chemical mechanical planarization (CMP) system is provided. The
system includes a polishing surface having a window defined
therethrough. A carrier configured to support a semiconductor
substrate is included. The carrier is further configured to force a
surface of the semiconductor substrate against the polishing
surface. An endpoint detector disposed below an underside of the
polishing surface is included. The endpoint detector is configured
to receive a reflected light signal from the surface of the
semiconductor substrate. The endpoint detector includes an emitter
for generating an intensity of light and a receiver for measuring a
reflected light spectrum from the intensity of light. The system
includes a controller in communication with the end point detector.
The controller is configured to generate a curve corresponding to
the reflected light spectrum by best fitting criterion.
It is to be understood that the foregoing general description and
the following detailed description are exemplary and explanatory
only and are not restrictive of the invention, as claimed.
BRIEF DESCRIPTION OF THE DRAWINGS
The accompanying drawings, which are incorporated in and constitute
part of this specification, illustrate exemplary embodiments of the
invention and together with the description serve to explain the
principles of the invention.
FIG. 1 is a graph of the wave number versus the normalized
amplitude for curves representing the reflectance spectrum and the
best fitting for the reflectance spectrum for a single shot in
accordance with one embodiment of the invention.
FIG. 2 is a simplified schematic diagram of a multilayer
transparent film structure on a semiconductor substrate in
accordance with one embodiment of the invention.
FIG. 3A and FIG. 3B are graphs illustrating spectra fitting for a
decreasing thickness during a CMP process in accordance with one
embodiment of the invention.
FIG. 4 is a graph representing the decrease of a film thickness
from shot to shot of an oxide layer that is less than about 600
nanometers thick in accordance with one embodiment of the
invention.
FIG. 5 is a simplified schematic diagram of a chemical mechanical
planarization (CMP) system having an end point detector (EPD) in
accordance with one embodiment of the invention.
FIG. 6 is an enlarged simplified schematic diagram of a cross
sectional view of a semiconductor substrate and the paths of light
rays in multiple reflections from the various layers disposed on
the semiconductor substrate in accordance with one embodiment of
the invention.
FIG. 7 is a flowchart of the method operations for determining a
thickness of a film disposed on a semiconductor substrate in
accordance with one embodiment of the invention.
DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
Several exemplary embodiments of the invention will now be
described in detail with reference to the accompanying drawings. It
should be appreciated that like numerals represent like
structures.
The embodiments described herein propose a method and apparatus
which allows for the reliable measurement the thickness of thin
films, which in turn can be used for endpoint determination. It
should be appreciated that while the embodiments described herein
allow for the measurement of films having a thickness that is less
than about 600 nanometers (nm), the embodiments are also effective
for measuring films having a thickness of 600 nm or more. In one
embodiment, the method and apparatus use the dominant reflectance
of silicon to determine a trench depth in a silicon substrate as
the upper layers are planarized or etched.
Processes, such as chemical mechanical planarization (CMP) or etch
processes deal not with a single spectrum but with series of
spectra corresponding to the decreasing thickness of the upper
layer. These spectra could be used together to extract low
components from reflected spectra signal corresponding to layers of
interest, such as an upper layer in one embodiment. Additionally,
the spectra can be used together to decrease system noise, decrease
the number of parameters in the fitting model and consequently
increase reliability of measurement results.
The reflectance spectra R(v) consists of N reflectances measured at
different wave numbers v.sub.i, i=1,2, . . . N and v.sub.1
<V.sub.2 < . . . <V.sub.N, thus
R(v)={R(v.sub.1), R(V.sub.2), . . . R(v.sub.N)}, where wave number
v=1/.lambda., .lambda. is a wave length.
The trench depth measurement presents reflectance R in a form
dependent on one parameter d.sub.0, which is the distance between
the top and the bottom of a trench in a Si substrate. In one
embodiment, the dominant reflecting surfaces is the trench in the
Si substrate. This property is mathematically represented by the
function: R(v,d.sub.0)=c+b*cos(2.pi.*2n.sub.0 *d.sub.0
*v)+.epsilon.(v)
where n.sub.0 is the refractive index of silicon oxide filled up
the trench, b and c are coefficients, and .epsilon.(v) is a small
contribution of other layers' interference to the reflectance
spectra.
Assuming .epsilon.(v) is negligible, then the real reflectance
spectra can be expressed as
R(v,d.sub.0).apprxeq.c+b*cos(2.pi.*2n.sub.0 *d.sub.0 *v).
Considering C.apprxeq..SIGMA..sub.i=1, . . . , N
R(v.sub.i,d.sub.0)/N, which is the average reflectance level.
Coefficient b depends on do and can be mathematically represented
as b={.SIGMA..sub.i=1, . . . , N [R(v.sub.i,d.sub.0)-c].sup.2
}/{.SIGMA..sub.i=1, . . . , N cos.sup.2 (2.pi.*2.sub.n0 *d.sub.0
*v.sub.i)}. Parameter d.sub.0 is found from the best fitting of
R(v,d.sub.0) according to minimum squares criterion.
FIG. 1 is a graph of the wave number versus the normalized
reflectance spectrum and the best fitting for the reflectance
spectrum for a single shot in accordance with one embodiment of the
invention. Solid line 100 represents the reflectance spectrum from
shot 11 and dotted line 102 represents the best fitting to solid
line 100. As will be explained in more detail below, the trench
depth is easily ascertained once the best fitting is identified,
since the multiple parameters needed to determine the trench depth
are all found from the best fitting curve. The trench depth
associated with the best fitting curve of FIG. 1 is 227 nm.
Consequently, the trench depth (d.sub.trench)=d.sub.0 =227 nm
provides an adequate estimation of the trench depth.
It should be appreciated that the trench depth (do) stays the same
for each shot as the layers above the trench are being removed as
shown with reference to FIG. 2. The intensity of the reflectance
spectra is strong, as the silicon layer is not transparent as the
films disposed over the silicon substrate can be. The reflected
spectrum is plotted as shown in FIG. 1 and a known cosine based
curve is fitted to find a best fit curve. The known cosine based
curve is associated with a trench depth that approximates the
trench depth for the reflected spectrum. Thus, the trench depth is
ascertained from one shot in this embodiment.
In one embodiment of the invention, the difficulties related to
restoration of the thickness of layers from the reflectance spectra
is overcome by considering the differences between spectra of
different shots rather than restoring the spectra themselves. In
this manner, the contribution that does not change from shot to
shot is discarded, such as interference between inner layer
boundaries. The information of interest, such as upper layer
thickness can be extracted by analyzing the difference between
spectra of different shots as described in the mathematical models
below. As a result of analyzing the differences, the number of
parameters in the mathematical model is significantly decreased and
the part of spectra corresponding to the upper layer can be
extracted.
FIG. 2 is a simplified schematic diagram of a multilayer
transparent film structure on a semiconductor substrate in
accordance with one embodiment of the invention. Layer 140 is
disposed over layer 142. Layer 142 is disposed over substrate 124.
In one embodiment, layers 140 and 142 are transparent films. Trench
143 is defined within semiconductor substrate 124. In one
embodiment, with respect to STI technology, layer 140 is an oxide
layer, such as silicon oxide and layer 142 is a nitride layer, such
as silicon nitride. As is generally known with STI technology, top
layer 140 is planarized along with a portion of layer 142.
Therefore, the trench depth, do, does not change. Thus, after a
planarization process a portion of layer 142, depicted by line 141,
remains. As mentioned above, the contribution to the spectra from
interference between surfaces 2 and 3 can be removed by calculating
the difference between two shots taken at different time points
during the planarization process as layer 140 is being removed and
as layer 142 is partially removed.
The process of simplifying the mathematical model when considering
the difference between two shots during the planarization or etch
process is discussed below. While the process is explained for any
number of layers on a substrate, reference can be made to FIG. 2,
which depicts a substrate having a step structure, i.e., layers 140
over 142; over substrate 124; and layer 140 filled into trench 143
of substrate 124. Distances in the function below, denoted by
d.sub.1, d.sub.2, . . . , d.sub.k, represent the distances between
upper surface 1 and 2, 3, . . . ,k+1-th underlying surfaces.
Indices of refraction denoted by n.sub.1,n.sub.2, . . . ,n.sub.k
--and correspond to the 1st, 2nd, . . . ,k-th layer, respectively.
Thus, reflectance R is mathematically represented by:
The term a.sub.0, which does not change with decreasing thickness
of the upper layer, includes average reflectance level and all
terms corresponding to interference between inner boundaries. The
other terms of the above function are interference terms between
the upper surface and underlying boundaries, where a.sub.1,
a.sub.2, . . . , a.sub.k are coefficients representing the weight
of the corresponding interference term. One skilled in the art will
appreciate that v is the wave number and v=1/.lambda., where
.lambda. is the wavelength.
The expression 1.sub.j =n.sub.1 *d.sub.1 +.alpha..sub.j (j=1, . . .
,k) of the above function represents the optical path, 1.sub.j,
between the upper surface and underlying j-th surface.
For instance, the optical path 1.sub.2 (FIG. 2) is equal to:
therefore, .alpha..sub.2 =n.sub.2 (d.sub.2 -d.sub.1).
If, in a process such as CMP, the thickness of the upper layer
reduces by a value of .DELTA. then reflectance R changes to
expression:
It should be appreciated that changes in values of coefficients
related to a change in absorption of the upper layer are negligible
and are disregarded. The spectra difference then becomes:
##EQU1##
The term a.sub.0, which is suppressing part of the reflectance and
contains some of the unknown characteristics of the inner layers,
cancels when the differences are considered. The spectra difference
can be transformed to a more convenient form using the
trigonometric identity cos .alpha.-cos .beta.=2 sin
{(.beta.-.alpha.)/2}*sin {(.alpha.+.beta.)/2}: ##EQU2##
The removed thickness, .DELTA., between close shots is small. That
is, .DELTA.<<d.sub.1.
Therefore, S(v,d.sub.1) is reduced to the expression: ##EQU3##
Due to present system configuration, factor 2*sin
(2*.pi.*n.sub.1,*.DELTA.*v) does not have practical impact on the
algorithm output, and will be disregarded in further
discussion.
It should be appreciated that the above function has less terms and
can be effectively used for fitting real reflectance spectra to
obtain the thickness of film layers in the structure. The fitting
function, i.e., the above expression for S(v,d.sub.1) has two
k-parameters a.sub.j and .alpha..sub.j, where j=1, . . . ,k, that
are common for all shots until the time when layer 140 is removed
completely.
For exemplary purposes, the efficiency of using the spectra
difference method described herein is shown on the thickness
measurement of the upper layers of a STI structure in CMP process.
For this example, the STI structure includes four surfaces and, in
turn, spectra difference S(v, d.sub.1) includes interference of the
upper surface with three underlying surfaces, i.e., k=3. To
approximate the spectra difference the above function is simplified
to the form:
where .alpha..sub.1 =0; .alpha..sub.2 =n.sub.2 *(d.sub.2 -d.sub.1);
.alpha..sub.3 =n.sub.1 *(d.sub.3 -d.sub.1). Amplitudes of terms are
considered equal, i.e., a.sub.1 =a.sub.1, =a.sub.2 =a.sub.3, so a
is normalizing coefficient. As a matter of fact, measuring the
trench oxide during CMP removal is of interest to determine the end
point of the CMP removal. Therefore, the above expression is
rewritten as a function of trench oxide-d.sub.3. Denoting the
thickness of layer 142 as d.sub.N, we obtain:
Continuing with the above example the spectra difference is
considered between shots j+h and shot j with h=3: S.sub.j
=R.sub.j+h -R.sub.j, where j=1, 2, . . . that is, the number of the
shot.
FIG. 3A and FIG. 3B are graphs illustrating spectra fitting for a
decreasing thickness during a CMP process in accordance with one
embodiment of the invention.
FIG. 3A represents the spectra difference using the above described
function between shot 10 and shot 7 of one exemplary process. Curve
104 of FIG. 3A is a real reflectance spectra difference between
shots 10 and 7. Curve 106 of FIG. 3A is a best fitting to real
spectra, from which trench oxide thickness d.sub.3 has been
calculated, where d.sub.3=604 nm.
For FIG. 3B, Curve 108 is a real reflectance spectra difference
between shots 49 and 46. Curve 110 of FIG. 3B is a best fitting to
real spectra, from which trench oxide thickness d.sub.3 has been
calculated where d.sub.3 =429 nm. It should be appreciated that
curves 106 and 110 are generated by a controller executing a best
fitting algorithm in one embodiment of the invention.
FIG. 4 is a graph representing the decrease of a film thickness
from shot to shot of an oxide layer that is less than about 600
nanometers thick in accordance with one embodiment of the
invention. It should be appreciated that the thickness of the film
of the semiconductor substrate is being decreased by a layer
removal process, such as a CMP or an etch process. Thus, the
substantially linear reduction of the thickness of a thin
transparent film, such as an oxide layer is captured.
FIG. 5 is a simplified schematic diagram of a chemical mechanical
planarization (CMP) system having an end point detector (EPD) in
accordance with one embodiment of the invention. CMP system 120
includes wafer carrier 122 supporting semiconductor substrate 124,
which is disposed over CMP belt 128. Air bearing platen 132 is
located on the underside of CMP belt 128. EPD 134 is located under
air bearing platen 132 such that an optical pathway exists between
EPD 134 through CMP belt 128 to semiconductor substrate 124. One
skilled in the art will appreciate that EPD 134 includes an emitter
for generating a light intensity and a receiver for receiving
reflected light reflected spectra from the generated light. The
optical pathway is enabled once every rotation of CMP belt 128
through window 126 disposed within CMP belt 128. CMP belt 128 is
driven around rollers 130. EPD 134 is in communication with
computer 136. It should be appreciated that while CMP system 120 is
shown as a belt system, the invention described herein can also be
applied to a rotary CMP system. Furthermore, EPD 134 can also be
incorporated in other semiconductor manufacturing operations
requiring thickness measurement, such as etch applications.
Still referring to FIG. 5, in one embodiment, once per rotation of
belt 128 EPD 134 takes a shot of the wafer surface being planarized
through window 126. It should be appreciated that wafer and
semiconductor substrate are interchangeable, as used herein. As
mentioned with reference to FIG. 1, a single shot can determine the
trench depth. Alternatively, two shots can be used to determine a
film thickness of a film on the semiconductor substrate. The
spectra detected by EPD 134 is normalized and then the fitting
procedure is applied. The found fitting curve defines the variables
needed to determine the thickness of the trench at the point in
time when the shot was taken or the thickness of the film when the
difference between shots is used. In one embodiment, where the
difference between the two reflectant spectra is used to determine
a film thickness, the best fitting curve is found based on a slope
fitting criterion. That is, the algorithm for finding the best
fitting curve includes a slope fitting component. It should be
appreciated that computer 136 is configured to execute the
algorithm for finding the best fitting curve.
FIG. 6 is an enlarged simplified schematic diagram of a cross
sectional view of a semiconductor substrate and the paths of light
rays in multiple reflections from the various layers disposed on
the semiconductor substrate in accordance with one embodiment of
the invention. Layer 140 is disposed over layer 142, which in turn
is disposed over substrate 124. In one embodiment with respect to
shallow trench isolation (STI), layer 140 is a silicon oxide layer,
layer 142 is a silicon nitride layer disposed over silicon
substrate 124. Reflectant spectra 144b, 146b, 148b and 150b are
generated from incident light 144a, 146a, 148a and 150a,
respectively. Thus, each layer disposed on the silicon substrate
generates a portion of the reflectant spectra. However, the
reflectant spectra is dominated by reflectant 144b and 146b from
the silicon substrate. Thus, by comparing two shots taken at
different time points, the interference generated from light rays
from layer 140 having thickness d1, layer 142 having thickness
d2-d1 and layer 143 having thickness d3 is taken into
consideration, and the best fitting curve can be found. Therefore,
the trench oxide thickness represented by distance d3 can be
calculated from the differences between successive or different
shots according to the function discussed above, as thickness d3 is
decreasing from the planarization operation. It should be
appreciated that an algorithm comparing the curve of the actual
signal to a best fitting curve can be used here. In one embodiment,
once thickness d3 reaches a predetermined thickness, the process
can be stopped as the end point has been reached. It should be
appreciated that while the embodiments described herein refer to a
CMP system, the invention is also applicable to other semiconductor
manufacturing operations that require the thickness to be measured
such as etch applications.
FIG. 7 is a flowchart of the method operations for determining a
thickness of a film disposed on a semiconductor substrate in
accordance with one embodiment of the invention. The method
initiates with operation 160 where multiple layers on a
semiconductor substrate are provided. In one embodiment the
multiple layers are associated with shallow trench isolation (STI)
technology. The method then advances to operation 162 where two
reflectance spectra, i.e., two shots, are generated. The
reflectance spectra is generated from a light beam of a certain
intensity directed toward the surface of the semiconductor
substrate in one embodiment. Here, the light beam is reflected from
the surface of the multiple layers and the reflected spectra from
each of the multiple layers are captured by a detector. The method
then proceeds to operation 164 where a difference between the two
reflectance spectra is calculated. In one embodiment, the
difference between the two reflectance spectra allows for the
cancellation of unknown terms so that the functions discussed above
can be simplified. That is, the multiple parameters can be reduced
by using the difference of two shots when determining the thickness
of a film. In another embodiment, a trench depth can be determined
from one of the two reflectance spectra.
The method of FIG. 7 then moves to operation 166 where the
difference between the two reflectance spectra is used to define a
curve. In one embodiment, the curve is a sinusoidal curve. Here,
the sinusoidal curve is approximated by a sine based function which
can be represented by the following function ##EQU4##
The method then advances to operation 168 where the curve of the
actual reflectance spectra approximated by best fitting curve to
determine a film thickness. In one embodiment, the film thickness
is less than about 600 nm. It should be appreciated that the
parameters associated with best fit curves have been identified.
Accordingly, once the best fitting curve has been identified the
film thickness is known. In one embodiment, a general purpose
computer can execute the algorithm.
In another embodiment of the invention, one of the two reflectance
spectra generated in operation 162 can be used to determine a
trench depth as described above with respect to FIG. 1. Here, one
of the two reflectance spectra is used as a real reflectance
spectrum. The real spectrum curve is then fitted by best fitting
solution using minimum squares criterion. As discussed above with
respect to the found sine based curves, the parameters associated
with each of the found cosine based curves have been identified.
Thus, once the best fitting cosine based curve has been identified,
the trench depth can be approximated. It should be appreciated that
only one shot is necessary here. In one embodiment, a cosine based
function generates the values for the cosine based curve. For
example, the cosine based function for approximating the trench
depth is:
cos (2.pi.*2n.sub.1 *d.sub.0 *v) in one embodiment.
Furthermore, the method described herein can be applied to
determine an endpoint of a semiconductor fabrication process that
removes a layer or layers from the semiconductor substrate, such as
a CMP process, an etch process, etc.
In summary, the present invention provides for the determination of
an endpoint of a semiconductor fabrication process, such as a CMP
or an etch process, through the cancellation of interference
between inner layer boundaries by considering differences between
spectra of different shots rather than restoring the spectra. The
invention has been described herein in terms of several exemplary
embodiments. Other embodiments of the invention will be apparent to
those skilled in the art from consideration of the specification
and practice of the invention. The embodiments and preferred
features described above should be considered exemplary, with the
invention being defined by the appended claims.
* * * * *